tf.experimental.dtensor.initialize_tpu_system
Initialize the TPU devices.
tf.experimental.dtensor.initialize_tpu_system(
enable_coordination_service=False
)
This functions performs additional TPU related initialization after
calling dtensor.initialize_multi_client
to initialize multi-client DTensor.
Refer to dtensor.initialize_multi_client
for relevant environment
variables that controls the initialization of multi-client DTensor.
Args |
enable_coordination_service
|
If true, enable distributed coordination
service to make sure that workers know the devices on each other, a
prerequisite for data transfer through cross-worker rendezvous.
|
Raises |
RuntimeError
|
If running inside a tf.function.
|
NotFoundError
|
If no TPU devices found in eager mode.
|
Except as otherwise noted, the content of this page is licensed under the Creative Commons Attribution 4.0 License, and code samples are licensed under the Apache 2.0 License. For details, see the Google Developers Site Policies. Java is a registered trademark of Oracle and/or its affiliates. Some content is licensed under the numpy license.
Last updated 2023-03-17 UTC.
[null,null,["Last updated 2023-03-17 UTC."],[],[]]